Data Mining for Studying the Impact of Reflection on Learning

Abstract

Title: Data Mining for Studying the Impact of Reflection on Learning Keywords: educational data mining, Reflect, learning behaviour, impact Abstract On-line Web-based education learning systems generate a large amount of students' log data and profiles that could be useful for educators and students. Hence, data mining techniques that enable the extraction of hidden and potentially useful information in educational databases have been employed to explore educational data. A new promising area of research called educational data mining (EDM) has emerged. Reflect is a Web-based learning system that supports learning by reflection. Reflection is a process in which individuals explore their experiences in order to gain new understanding and appreciation, and research suggests that reflection improves learning. The Reflect system has been used at the University of Sydney’s School of Information Technology for several years as a source of learning and practice in addition to the classroom teaching. Using the data from a system that promotes reflection for learning (such as the Reflect system), this thesis focuses on the investigation of how reflection helps students in their learning. The main objective is to study students' learning behaviour associated with positive and negative outcomes (in exams) by utilising data mining techniques to search for previously unknown, potentially useful hidden information in the database. The approach in this study was, first, to explore the data by means of statistical analyses. Then, popular data mining algorithms such as the K-means and J48 algorithms were utilised to cluster and classify students according to their learning behaviours in using Reflect. The Apriori algorithm was also employed to find associations among the data attributes that lead to success. We were able to group and classify students according to their activities in the Reflect system, and identified some activities associated with student performance and learning outcomes (high, moderate or low exam marks). We concluded that the approach resulted in the identification of some learning behaviours that have important impacts on student performance.